The proportion of individuals with advanced chronic kidney disease who undergo chronic peritoneal dialysis (CPD) treatment, currently <8% of the 400,000 dialysis patients in the USA, has been declining steadily in the past several years. However, the total dialysis patient population continues to grow fast and consumes >6% of the Medicare budget. Recent reports indicating inferior survival in certain groups of CPD patients compared to their maintenance hemodialysis (MHD) counterparts may have played a major role in the CPD underutilization leading to over-utilization of the more expensive MHD. Moreover, the associations between cardiovascular risk factors and outcome in both MHD and CPD patients appear to be somewhat different than the general population. There are inconsistent or even contradictory outcome data when CPD and MHD patients are compared in observational studies. Over-adjustment for inherent advantages of CPD treatment such as slower drop in residual renal function, inadequate analyses of potential interactions or inability to perform subgroup analyses due to small sample size, non-existent or suboptimal CPD to MHD matching in comparative studies, and lack of utilization of novel epidemiologic techniques such as causal structural models, propensity score and instrumental variable, which better adjust for bias by indication, can be among the reasons for discrepant survival data in CPD patients. We propose to study the 5-year (7/2001-6/2006) data of ~17,000 CPD patients in a large dialysis provider, consisting of ~10,000 newly started CPD patients. We will use incidence propensity based density case-control design to match each incident CPD patient to two MHD patients with similar demographic and comorbidity constellations in the same center and at the same time;and will develop and examine several novel epidemiologic models and compare them to traditional survival models. We plan to examine following questions: (1) Is the CPD survival inferior to MHD in all subgroups of dialysis patients, or are there distinct subgroups of patients who may benefit more from either dialysis modalities? And can causal structural models, propensity score and instrumental variable, compared to traditional survival models, better identify subgroups of CPD patients with superior survival? (2) Are the associations between risk factors and outcomes similar in both 10,000 CPD patients and their 20,000 propensity-based incidence- density matched MHD patients and are these associations different than what is observed in the entire 200,000 MHD cohort, and can competing risks or other confounding factors explain the discrepant outcome data? (3) Can the created CPD cohort and developed methodological techniques be used in future studies related to CPD practice and outcome as well as other CKD population related questions? PUBLIC HEALTH RELEVANCE: Examining survival data in dialysis patient databases may lead to identification of subgroups of dialysis patients with superior survival with either dialysis modalities, leading to better clinical and economic outcomes. Studying administrative databases by developing and combining novel epidemiologic models to better control for bias by medical indication may result in developing practical methods to better examine outcomes in observational cohorts.